Healthcare has thousands of AI tools competing for attention, and no one has the time (or expertise) to sort through them. A new kind of middleman is coming: the AI broker. Just like insurance brokers navigate plans for beneficiaries and PBMs navigate formularies for payers, AI brokers will navigate the vendor landscape for healthcare purchasers. We don't fully have them yet. But we will.

Every time healthcare gets complicated enough, a middleman appears to absorb the complexity.

When drug pricing got too convoluted for insurers to manage, pharmacy benefit managers stepped in. When insurance plans multiplied beyond what individual beneficiaries could evaluate, insurance brokers emerged to match people to coverage. When hospitals couldn't keep up with billing complexity, revenue cycle management companies took over the back office.

Now we're watching the same pattern unfold in real time… just with artificial intelligence.

The healthcare AI market has exploded. There are thousands of tools available today: ambient documentation, prior authorization automation, clinical decision support, RCM optimization, scheduling agents, population health analytics, imaging AI, and more. Every health system, every large practice group, every payer faces the same overwhelming question: which tools do we actually buy, how do we evaluate them, and how do we fit them together?

Most organizations don't have the internal bandwidth to answer that question well. And that gap is exactly where a new middleman will plant its flag: the AI broker.

The AI broker would sit between the purchaser and the AI vendor ecosystem.

A hospital administrator walks in and says: I need an agentic AI solution for scheduling. I need something that can manage a patient's care journey through the hospital. I also need an RCM tool that reduces claim denials. The broker listens, assesses the organization's existing tech stack, budget constraints, and workflow needs—and then goes out and finds the best-fit options.

In some cases, the broker might even assemble a quasi-integrated solution: matching multiple point solutions together like puzzle pieces, so that the purchaser gets something that feels like a unified system without having to negotiate five separate enterprise contracts. Think of it as a general contractor for healthcare AI. While they don't build the tools, they design the stack and manage the build.

Right now, the closest thing we have to this model is Elion, a health IT intelligence platform that catalogs thousands of healthcare AI products by category and use case. Elion raised $9.3 million in a seed round backed by NEA and Cedars Sinai Health Ventures, and over 60% of U.S. health systems are already using the platform. They provide the research, the market maps, the buyer's guides. What they don't do (at least not yet) is make the decision for you. The onus still sits with the purchaser.

That's the gap. The AI broker fills it by going from intelligence platform to active advisor—someone who takes on the evaluation and matching work rather than just surfacing the data.

Optum has launched its own AI Marketplace. Autonomize AI released a Healthcare Agents Marketplace with over 100 pre-built tools. Advisory firms like ReMedi Health Solutions have started offering AI vendor selection services for health systems. The scaffolding is being built. The full broker model—with active matching, curated stacking, and economic skin in the game—is the next logical step.

To understand where this is going, it helps to trace how we got here.

Diagnostic imaging AI was the first major AI solution in medicine. Radiology tools that could flag findings on mammograms and chest CTs arrived before most of us knew what a large language model was. The use cases were narrow, the evidence was accumulating, and the procurement decisions were relatively contained (one department, one workflow, one vendor).

Then ChatGPT launched in late 2022, and the use case universe cracked wide open.

Within eighteen months, every vendor in healthcare had attached "AI" to their pitch deck. Ambient documentation tools went from niche to mainstream almost overnight. Clinical decision support, care gap automation, prior auth bots, and patient engagement agents followed rapidly. By 2024 and 2025, health system CIOs were staring at a landscape with hundreds of credible tools and no clean way to evaluate them.

The market has grown and is fragmented. Different tools work better for different EHR environments, different patient populations, different organizational sizes. A tool that works at Mass General may be the wrong fit for a 200-bed regional hospital in the Midwest. The complexity of matching tools to context is the same kind of complexity that birthed the PBM and the insurance broker. And we know how that story goes.

The business model almost writes itself.

An insurance broker earns a commission (typically a percentage of the premium) from the insurer each time they place a client. The broker's incentive is aligned, at least in theory, with finding the right match because their reputation depends on it, even if the financial incentive can introduce conflicts.

A PBM captures value through spread pricing, rebates, and administrative fees—sitting in the middle and monetizing every transaction that flows through it.

An AI broker would likely operate somewhere between these two models. Options include:

The tension, as with PBMs, will come down to incentive alignment. If the broker is paid by the vendors, they'll be tempted to steer toward the vendors who pay the most. If they're paid by the purchaser, they have a cleaner fiduciary relationship—but the model may be harder to scale. Group Purchasing Organizations have an interesting model like this.

We've seen this exact debate play out before. It doesn't end cleanly. But it also doesn't stop the middleman from emerging.

As for the healthcare system broadly, the AI market is already showing signs of consolidation fatigue. Health systems are wary of vendor sprawl, integration debt, and shiny tools that don't deliver on their promises. An AI broker with a track record of honest evaluation could actually improve the quality of AI adoption across the industry, accelerating tools that work and filtering out the ones that don't. Or, if the incentives go sideways, it becomes another toll booth. Another entity collecting a fee for a service that only exists because we built a system too complicated to navigate without one.

For now, yes. Most health systems are managing by cobbling together internal IT teams, consultants, and platforms like Elion to make their AI decisions. The process is slow and imperfect, but it functions.

The question isn't whether we need an AI broker today. The question is whether we'll be able to avoid one in five years as the market grows more complex and the stakes get higher.

I don't think we will. The pattern is too familiar. Healthcare has never been able to resist the emergence of a middleman when complexity outpaces capacity. And right now, the AI landscape in healthcare is doing exactly that.

The AI broker is coming. The more important question is whether we build the model with the right incentives from the start—or whether we spend the next decade complaining about it the way we complain about PBMs.

In summary, the AI broker is a prediction. I don’t want it to sound like it’s a warning. Middlemen emerge in healthcare because the underlying system demands them. When that happens, the best we can do as physicians is pay attention early before the incentive structures calcify and the middleman becomes too entrenched to question.

If you work at a health system and you're evaluating AI vendors right now, push for transparency. Ask consultants and advisory firms who's paying them. Ask vendors for independent clinical validation data, not just case studies they commissioned themselves. And if someone pitches you a bundled AI stack, ask whether the bundle reflects your needs—or their margin.

The tools are powerful. The brokers, when they arrive, will be necessary. Just make sure you know whose side they're on.

Read more from The Middlemen Series here.